Why does Japanese not use spaces between words?

Because the writing system already marks boundaries a different way. Japanese mixes three scripts in one sentence - kanji for content words, hiragana for grammatical endings and particles, katakana for loanwords - in a system called 漢字仮名交じり文, kanji-kana mixed writing. Every switch from one script to another signals a boundary, so a space would be redundant. Texts that genuinely lack this signal, such as books for young children written entirely in hiragana, do add spaces; that convention is called 分かち書き.

How do Japanese readers know where a word ends?

Mostly by script change plus a small set of grammatical markers. A run of kanji opens a content word. The hiragana immediately after it is okurigana and belongs to the same word. Eleven single-character particles - はがをにでへとものかね - act as fences between phrases. A run of katakana is a single island. And a closed set of endings (ます, ました, ません, です, ている, たい, ない) closes a word. Together these five signals resolve the overwhelming majority of sentences without any conscious effort.

Is it true that kanji make Japanese easier to read, not harder?

For continuous text, yes. Kanji visually chunk a sentence into content words, so the eye can locate meaning without decoding every syllable. The classic demonstration is ここではきものをぬいでください, which in pure hiragana can be read as ここで・はきもの (take off your footwear here) or ここでは・きもの (take off your kimono here). Written with kanji as ここで履物を脱いでください or ここでは着物を脱いでください, the ambiguity disappears completely. This is also why a sentence full of kanji feels slow to a beginner but fast to anyone who knows the characters.

Which particles act as word boundaries in Japanese?

The high-frequency single-character particles are は (topic, pronounced wa), が (subject), を (object, pronounced o), に (destination or time), で (location of action or means), へ (direction, pronounced e), と (and, with), も (also), の (possessive or linking), か (question) and ね (confirmation). They can also appear in compound form - には, では, との, からの, にも - which still marks a boundary. The main trap is that は, が, で, に and と also occur inside words, so a single hiragana is only a fence when a complete content word sits in front of it.

What is the fastest way to practise segmenting Japanese?

Work on short real sentences rather than long passages, and check your split immediately so a wrong boundary never sets. A tool that segments a sentence into individual words and shows which ones you already know turns this from guesswork into feedback: reading real captions or a scanned page with the boundaries drawn in trains the eye far faster than reading graded text with no answer key. Five to ten sentences a day for two weeks is usually enough for the rules below to stop being conscious.

Japanese Has No Spaces: How to Tell Where Words Start and End

The short version: Japanese does not need spaces because the change of script is the space. Content words wear kanji or katakana; grammar wears hiragana. Five signals do the rest: kanji blocks open words, hiragana tails stay attached, eleven single-character particles act as fences, katakana runs are single islands, and a small closed set of endings closes the word.

Level: N5–N4. Every example sentence uses beginner grammar and has furigana on every kanji, full hiragana underneath and a complete English translation. You do not need to know the vocabulary to follow the method.

There is a specific moment every learner hits, usually somewhere in the second or third month. You have done hiragana. You have done katakana. You have a hundred kanji and a page of grammar notes. Then you open something real — a sign, a manga panel, a caption — and the sentence lands as a single uninterrupted string:

わたしはまいにちでんしゃでがっこうへいきます

Nothing is unknown. Every syllable is one you can read aloud. And still it does not resolve, because your eyes have nowhere to stop. The problem is not vocabulary. It is that nobody ever taught you where a Japanese word ends.

Why Japanese can get away with no spaces

English marks word boundaries with a blank. Japanese marks them with a change of script. A normal Japanese sentence is written in 漢字仮名交かんじかなまじり文ぶん — “kanji-kana mixed writing” — and the mixing is not decoration. It is a division of labour:

Script	Carries	Examples
Kanji	Content: nouns, verb stems, adjective stems — the meaning	電車でんしゃ, 学校がっこう, 食たべ, 高たか
Hiragana	Grammar: particles, endings, inflection — the glue	は, を, に, ます, でした, ている
Katakana	Loanwords, names, emphasis — imported things	コンビニ, テレビ, アルバイト

Once you know that, the wall of text stops being a wall. Every switch from hiragana to kanji is the start of a new word. Every switch from kanji back to hiragana is the start of its tail or the next particle. The sentence is already spaced — just in a code you had not been given.

Here is the same sentence written the way it is actually written, in real Japanese:

The same sentence, with kanji

わたしはまいにちでんしゃでがっこうへいきます

私わたしは毎日まいにち電車でんしゃで学校がっこうへ行いきます。

I go to school by train every day.

Nothing was added except kanji — and the sentence now has six visible chunks instead of one. This is the entire reason Japanese keeps kanji.

content word (kanji)particle — a fencegrammar / endingkatakana island

The five rules

Rule 1 · A run of kanji opens a word

One to three kanji in a row is almost always one word. 電車でんしゃ is a word. 日本語にほんご is a word. 図書館としょかん is a word. Longer runs do occur in formal writing, but at N5–N4 you can treat a kanji run as a single unit and be right nearly every time.

Do this first. Mark the kanji blocks and ignore everything else — those are your anchors, and once they are placed, the hiragana between them has almost nowhere to hide.

Rule 2 · The hiragana tail belongs to the kanji in front of it

Hiragana sitting directly behind a kanji is 送おくり仮名がな — okurigana — the inflecting tail of that same word. 食たべます is one word, not 食 followed by べます. 大おおきい is one word. 話はなしました is one word.

So: after a kanji, keep attaching hiragana until you hit a particle or another kanji. That is the boundary.

Rule 3 · Eleven single hiragana are fences

These do the job spaces do in English. When one appears after a complete content word, there is a boundary on both sides of it:

はがをにでへとものかね

They also stack into compound fences — には, では, との, からの, にも — which still mark a boundary. Learn to see these eleven and roughly half of all segmentation solves itself. The full behaviour of each one is in the particles guide.

Rule 4 · A katakana run is one island

An unbroken stretch of katakana is nearly always a single word: コンビニ, アルバイト, エスカレーター. The script change marks both ends for you at once, which makes katakana the easiest boundary in the language to spot — and, awkwardly, the hardest script for beginners to actually read. If katakana slows you down, that is a separate and very common problem: see katakana blindness.

The one exception: a middle dot ・ splits a katakana run into parts, usually a foreign first and last name — ジョン・スミス.

Rule 5 · Endings are a closed set, and they close the word

You do not need to parse the tail character by character. Japanese endings come from a small, fixed inventory, and recognising the shape is enough:

ます · ました · ません · ませんでした · です · でした · ている · ています · たい · ない · なかった · ください

When you reach one of these, the word ends there. In a main clause, so usually does the sentence — which is also why Japanese listening feels like waiting: the verb, the negation and the tense all arrive at the very end.

Eight sentences, dissected

Read the grey line first — that is the sentence with no boundaries. Try to split it, then look at the coloured line.

1 · two fences, three words

わたしはがくせいです

私わたしは学生がくせいです。

I am a student.

2 · katakana island in the middle

コンビニでおべんとうをかいました

コンビニでお弁当べんとうを買かいました。

I bought a bento at the convenience store.

Notice お弁当. The polite prefix お is hiragana but belongs to the word in front of it, not behind it. お is the one hiragana that attaches forwards — お金かね, お茶ちゃ, お名前なまえ.

3 · two fences stacked

きょうはあめがふっています

今日きょうは雨あめが降ふっています。

It is raining today.

降っています is one chunk, not three. て-form plus います is a single grammatical unit — a state in progress. Splitting it is the most common beginner error at this level.

4 · the の linker

これはわたしのともだちのほんです

これは私わたしの友ともだちの本ほんです。

This is my friend's book.

Two の, two boundaries. Each の glues a noun to the next one, and reading right to left gives you the head: a book → of a friend → of mine.

5 · a long tail that is still one word

きのうはがっこうにいきませんでした

昨日きのうは学校がっこうに行いきませんでした。

I did not go to school yesterday.

行きませんでした is eight kana and one word. Long tails frighten beginners, but the tail is from a closed set — ません + でした — so there is nothing to parse, only to recognise.

6 · a compound fence

つくえのうえにはなにもありません

机つくえの上うえにはなにもありません。

There is nothing on the desk.

には is one fence, not two words — に marking the location and は adding contrast. And the string なにも is a question word plus も, which with a negative verb means nothing at all.

7 · N4: two clauses joined by て

あさごはんをたべてからでんしゃにのります

朝あさごはんを食たべてから電車でんしゃに乗のります。

After eating breakfast, I get on the train.

から here is not the particle “from”. Attached to a て-form it means after doing. Same two characters, different job — which is why the fence rule only fires when a complete content word sits in front.

8 · N4: a relative clause, no comma, no space

きのうかったほんはとてもおもしろかったです

昨日きのう買かった本ほんはとても面白おもしろかったです。

The book I bought yesterday was very interesting.

買った本 is a relative clause — “the book that I bought” — and Japanese marks it with nothing at all: no that, no which, no comma. A plain-form verb sitting immediately before a noun is modifying it. This single pattern is the biggest reading jump between N5 and N4.

The sentence that proves why kanji exist

Here is the classic demonstration, and it is worth sitting with for a minute. Written entirely in hiragana, this sentence has two completely different readings:

ここではきものをぬいでください

Split A

ここで·履物はきものを·脱ぬいでくださいHere · footwear · please take off.
“Please take your shoes off here.”

Split B

ここでは·着物きものを·脱ぬいでくださいIn this place · kimono · please take off.
“Please take your kimono off here.”

Same kana, same order, one character of difference in where the boundary falls — and a genuinely awkward misunderstanding. In real written Japanese this ambiguity cannot occur, because the kanji fix the boundaries before your eye even arrives.

There is a shorter one that does the same job: くるまでまとう is either 車くるまで待まとう (let's wait in the car) or 来くるまで待まとう (let's wait until they come).

The reframe that changes everything. Beginners experience kanji as the obstacle between them and reading. It is the opposite: kanji are the spaces. Learning them is not extra work piled on top of reading — it is the thing that makes reading possible at speed. This is why learners who push through the first two hundred characters report that Japanese suddenly “looks like text” rather than like noise.

Three traps that survive the rules

Trap 1: a single hiragana that is not a particle. は, が, で, に and と all appear inside words too — 初はじめ, 外国がいこく, できます, 日曜日にちようび. The fence rule only fires when a complete content word sits in front of the character. If what precedes it is a fragment, you have split in the wrong place.

Trap 2: では and でも. では can be a compound particle (as for at/in) or the start of ではない (is not). でも can be but at the head of a sentence, or even attached to a noun. Read on one more chunk before committing — in Japanese, ambiguity is almost always resolved by what comes after, not before.

Trap 3: the all-hiragana stretch. Casual writing, children's books, chat messages and some signage drop kanji entirely, and there the rules genuinely thin out. This is the one situation where native readers slow down too. Children's books compensate by adding real spaces — a convention called 分わかち書がき.

How to practise this so it becomes automatic

Segmentation is a perceptual skill, not a knowledge one. You cannot study your way into it; you have to see enough correct splits that your eye starts predicting them. Three conditions make that happen fast:

Short sentences, not passages. Ten dissected sentences teach more than one page read anxiously. Immediate feedback. A wrong boundary that goes uncorrected sets like concrete — you need to see the right answer within seconds, not at the end of a chapter. Real material. Graded readers are already gently spaced by their simple vocabulary; real captions, signs and pages are not, which is exactly why they are the useful practice.

This is the one part of reading where the right tool genuinely changes the timeline. Pointing the OCR scanner at a real page or sign gives you the characters with readings and meanings attached instead of a wall to guess at. And learning from a video is the purest form of this drill: paste a link, and every spoken sentence is segmented into individual words with furigana, marked for which ones your own level already covers, with translation on demand — which is exactly the boundary practice above, on material you actually want to watch. Word segmentation there runs on-device, so it works line by line at whatever pace you read.

Quick reference

Why no spaces: Script change does the job. 漢字仮名交じり文 = kanji for content, hiragana for grammar, katakana for loanwords.
Rule 1: A run of 1–3 kanji is one content word. Mark these first.
Rule 2: Hiragana directly after kanji is okurigana — same word. Keep attaching until a particle or another kanji.
Rule 3: Fences: はがをにでへとものかね, plus compounds には, では, との, からの, にも.
Rule 4: A katakana run is one island. A middle dot ・ splits it (usually a foreign name).
Rule 5: Endings are a closed set: ます, ました, ません, ませんでした, です, でした, ている, たい, ない, なかった, ください.
The ambiguity proof: ここではきものをぬいでください = ここで・履物 (shoes) or ここでは・着物 (kimono). Kanji remove the ambiguity.
Spaces do exist: 分かち書き — spacing used in all-hiragana texts for young children, where the script signal is absent.

Frequently Asked Questions

Because the script mix already marks boundaries. Japanese writes content words in kanji or katakana and grammar in hiragana — 漢字仮名交じり文 — so every script switch signals an edge and a space would be redundant. Texts without that signal, such as all-hiragana books for young children, do add spaces; the convention is called 分かち書き.

Five signals: a kanji run opens a word; the hiragana behind it is okurigana and stays attached; eleven single-character particles (はがをにでへとものかね) act as fences; a katakana run is one island; and a closed set of endings (ます, ました, ません, です, ている, たい, ない) closes the word.

Harder to learn, far easier to read. ここではきものをぬいでください can mean “take off your shoes here” or “take off your kimono here” depending on the split; written with kanji, the ambiguity vanishes. Kanji chunk a sentence visually, which is why text stops looking like noise once you know a couple of hundred of them.

はがをにでへとものかね, plus compound forms には, では, との, からの, にも. The catch is that these same characters appear inside words (初め, 外国, できます), so treat one as a fence only when a complete content word sits in front of it.

Short real sentences with immediate feedback, five to ten a day. Seeing the correct split within seconds is what trains the eye; a wrong boundary left uncorrected sets hard. Reading material where every sentence is already split into words — captions, a scanned page with readings attached — beats guessing at graded text with no answer key.

See the boundaries drawn in, on real Japanese

Kanjijo turns unspaced Japanese into something you can actually read. Paste a link to any Japanese video and every sentence is split into individual words with furigana, flagged against your own level, with translation on demand. Point the OCR scanner at a sign, a menu or a manga page and get readings and meanings on the spot. Underneath sits the full N5–N1 kanji, vocabulary and grammar path with exclusive mnemonics for every kanji and every word, an SRS engine, home, lock screen and interactive test widgets, JLPT reading and listening practice, mock tests, conversation, shadowing and dictation. Free on iPhone.

Download Kanjijo Free